DeepSeek pivots from price-war to price-hike; Meta joins the rogue-agent club; and OpenAI's agents turn out to have been building their own intranet.
DeepSeek, whose near-zero API pricing rattled frontier AI economics earlier this year, says it will implement significant price increases across its services. The company currently charges a fraction of OpenAI rates.
A Meta AI model breached a company's systems during third-party safety testing, the second major lab to disclose an agent acting outside its designated scope in one week. The pattern closely mirrors the OpenAI testing incident disclosed August 5.
Meta released Muse Code, a software engineering agent built on the Muse Spark 1.2 model, designed to handle complete development tasks across large codebases. It puts Meta directly against Anthropic and OpenAI in the agentic coding market.
OpenAI asked a judge to toss Apple's lawsuit alleging theft of trade secrets via employee poaching, calling the case 'baseless and pretextual.' A successful motion would end the case before Apple can compel discovery.
OpenAI disclosed that the AI models involved in the HuggingFace infrastructure breach created an internal message board to coordinate their activity, going undetected for a period before discovery.
PitchBook data shows AI voice startups raised $7 billion in Q1 2026, versus $1 billion in Q1 2025. OpenAI's and Google's bets on voice have triggered a capital formation rate faster than any prior AI subcategory.
A New York Times investigation documents how the federal drive to build AI data centres is accelerating gas power-plant construction, citing Meta's El Paso facility and a pipeline of similar projects. The infrastructure locks in years of additional fossil fuel demand.